Overview

The Cockpit Copilot is a specialised AI chatbot developed by Nezasa for TripBuilder. This intelligent tool assists users with inquiries about itineraries available in TripBuilder Customer Care, making it easier to access information and manage itineraries after bookings.

Cockpit Copilot uses natural language processing, allowing users to communicate without rigid syntax or rules. It swiftly identifies and explains itinerary and component information, eliminating the need for Customer Care users to consult various TripBuilder pages.

 

Important: This feature is available for all our customers and requires Nezasa activation. Existing customers must opt-in and accept the terms of service before enabling the feature. Please contact our Support team or your customer success manager to configure it under Settings > General Settings > Cockpit > Cockpit Copilot AI. Four visibility levels are available for configuration (listed from less to more restrictive): Enabled for all Customer Care Users* Enabled for (Customer) admins and Nezasa Users Only Enabled for Nezasa Users Only Disabled *The main beneficiaries of the feature right now are itinerary owners rather than suppliers in Customer Care (due to historical information on itineraries currently being relevant only to the itinerary owner). The feature should not be enabled or used by customers/users handling operations in the supplying cockpit (i.e. module owner for Peer-to-Peer).

 

Terminology

Term	Explanation
AI	Artificial Intelligence (AI) refers to the simulation of human intelligence in machines programmed to think and learn like humans, performing tasks such as problem-solving, decision-making, and language understanding. The Cockpit Copilot is based on AI technology.
Chatbot	A chatbot is a software that simulates a conversation with human users, primarily online. Chatbots use AI to interpret and respond to user inputs in natural language. The Cockpit Copilot is a chatbot.
LLM	Large Language Model (LLM) is a type of AI designed to understand and generate human-like text. The Cockpit Copilot is based on LLM technology.
PII	Personally Identifiable Information (PII) refers to any data that can identify a particular individual. The Cockpit Copilot applies data anonymisation to PII when interfacing with external non-Nezasa software platforms.
Prompt	In AI and natural language processing, a prompt is an input or instruction given to an AI model to generate a response or complete a task. Users communicate with the Cockpit Copilot using prompts.

 

Value

The Cockpit Copilot significantly improves TripBuilder’s booking experience by streamlining information access and reducing friction in post-booking operations through natural language processing.

Travel itineraries are often complex and may undergo multiple changes, making key booking details scattered across various customer care pages or inaccessible via the Customer Care UI. The Cockpit Copilot addresses this complexity, aiding in understanding bookings, processing changes, and expediting business workflows.

Key benefits:

• Reduces manual tasks and streamlines the post-booking process for Tour Operators
• Speeds up access to information.
• Provides seamless booking insights, reducing support tickets and improving user experience.
• Increases operations efficiency
• Improves customer satisfaction.

 

General Usage Remarks

Interaction Scope and Process

The Cockpit Copilot operates within the scope of a single itinerary. This means:

• Users must select one itinerary that will serve as the context for interaction.
• Prompts related to multiple itineraries will not be answered.
• Prompts outside the scope of relevant itinerary questions will not be answered.
• The provided information will only refer to the selected itinerary.

The use of the chatbot is typically an iterative process where the Customer Care user enters different prompts to obtain the required information, either by adding details with each prompt or drilling down further. During an ongoing conversation between the Customer Care user and the Cockpit Copilot, the context of previous prompts and answers is maintained, allowing prompts to refer to the conversation’s context.

However, the conversation context is lost when the conversation window is closed. Reopening the Cockpit Copilot for an itinerary where a previous conversation took place will not restore the previous conversation context.

Please note: If you wish to continue a conversation with the Cockpit Copilot later, ask the chatbot for a summary at the end of the conversation. Store this information (e.g., as a text file) and provide it as a prompt when starting a new conversation to re-establish the previous context.

 

Multi-language Support

While this guide discusses interactions in English, the Cockpit Copilot supports multiple languages for both prompts and results. The language selection is independent of the user language set in the TripBuilder interface. Some TripBuilder terms may not be fully translated when using non-English languages.

 

Cockpit Copilot Walkthrough

Understanding the Cockpit Copilot's usage is best achieved through our screencasts. You can use the feature with suggested prompts or your own.

Please note: As the conversation window with Cockpit Copilot is presented on a full screen, we suggest using a different browser tab with the itinerary view in Customer Care for additional context.

 

General usage

Screencast available here

In this screencast:

• How to access the Cockpit Copilot within the scope of an itinerary.
• Asking the Cockpit Copilot for example prompts.
• Basic prompt interactions related to passengers' special preferences and meal plans included with booked accommodations.

Prompts used:

• Copilot, can you tell me what sort of prompts I can ask you to find out more information about my itinerary?
• Any special preferences from my passengers?
• Can you tell me what meals have been included in each hotel?

 

Cancellations

Screencast available here

In this screencast:

• Use the Cockpit Copilot to understand the cancellation conditions for itinerary components.
• Format information on itinerary components for data export purposes.
• Build a data table incrementally based on iterative prompting.
• Apply conditional logic between itinerary component dates and a given prompt date for data display.

Prompts used:

• Give me all the hotels in a table that is easy to export into Excel - using the columns hotel name and the start and end dates
• Now add the cancellation fee aplicable if I was to cancel the whole itinerary today
• Please add the first date some cancellation fees apply on this itinerary including the fee

 

Pricing

Screencast available here 

In this screencast:

• Use the Cockpit Copilot to understand how the sales price was derived.
  Drill down into the pricing elements with additional prompting (e.g., promo codes, discounts, fees paid at the destination).

Prompts used:

• I want to understand my sales price - can you tell me what fees/surcharges and discounts have been applied to the booking?
• Do you know more details about the promo code?
• And what about the 100 discount?
• And there are any fees that are paid in destination not included in the price?

Agency Commissions

Prompts used:

• I want to understand the commission that has been applied - give me a full breakdown and a total of commission applied to this itinerary. 

Examples of prompts with results 

Itinerary History 

Screencast available here 

 

Prompts to use:

• I'm a tour operator and I'm performing an audit on this itinerary, I need to understand all post booking changes after the initial booking was made including all the pricing and position changes. Please provide me the full information so I can do a good audit from it.

• Give me all the changes (including the positions) including price changes after the initial booking was made. 

• Put together a simple table allowing me to compare the changes made across all booking changes and cancellations. Use the Booking Changes as columns.

• I added the activity component (position) Discover 7 Monuments in Toledo with Wine Tasting, Madrid to the itinerary it cost 50 EUR extra, the itinerary sales price did not increase by this amount, can you explain to me why?

• What was the travel start date of the original itinerary?

Example use case with results: 

• (Prompt 1) Give me a full breakdown of all changes that occurred on the third booking change
• (Prompt 2) Give me it as a comparison to the second booking change
• (Prompt 3) Also include a new column for the first change/initial booking

• Times the keep sales price feature was used after the initial booking.

Example Customer Care Prompts

Prompts help users start exploring the Cockpit Copilot. Here are some examples:

Please note: For better organisation or data export purposes, specify in your prompt how the results should be formatted (e.g., tabular or CSV format).

 

Category	Use Case
General Inquiries	What questions can I ask? Copilot, can you give me a summary of the questions and answers I have asked and been given?
Basic Search and Reporting	What’s the status of this booking and what is the outstanding balance to be paid?
	I have to send a quick summary of the booking to our internal customer support team, can you put me something together?
	Copilot I’d like a simple table containing: Hotels on the itinerary Hotel name name Net Price Sales Price Simple one line description of the hotel Put that into a CSV format
	Can you represent this booking as an Excel file. Extracting data as a CSV
	Can I change the name of X product? (Finding out if I can edit the attribute of a component).
	Is anything incomplete/not booked on my itinerary? Itinerary status (Manual Processing, Change Initiated) Component status (Open, None)
Pricing and Fees	Copilot, can you give me a breakdown of all the fees and discounts applied to this itinerary? This could be: Processing Fees Discount Promo Code Yield Price Discount Tour Operator Surcharge Optional Service Charge Purchase Cancellation Fees Locally paid fees
	What is my profit with this booking? (Follow up prompt) Drill-down further please….
	Copilot, aside from the sum of the component sales prices, what else is contributing to the internal sales price?
	Is the processing fee factored into the sales price or the customer sales price?
	Copilot, do the open payments on the billing page match the customer sales price?
	Copilot, what is the difference between the customer sales price and the sales price?
Cancellation Rules and Fees	Copilot if I cancelled the Best Western Plus Ajaccio Amiraute Hotel & Residence , Ajaccio today, what would the cancellation fee for that component be?
	If I cancelled the booking today, what would the cancellation fee that I charge my customer be? Tell me what percentage that is?

 

Know more: If a prompt doesn’t give you an answer, don't give up! Try phrasing it differently or more clearly. The quality of the data retrieved will improve as the chatbot receives more context from the user.  For a better understanding of how prompting works and best practices, please refer to Prompt Engineering Guide.

 

Supported and Future Work/Limitations

This section provides additional information on which Customer Care areas are in the scope of the current version of Cockpit Copilot, and others that will be considered as future work.

Supported	Future Work (not supported)
Prompts for a single itinerary	Prompts for multiple itineraries Prompts that correlate itineraries or information in itineraries
Main prompting categories: General information on an itinerary Itinerary and component history. Components Travel Documentation Pricing/Fees/Discounts Billing Cancellation Rules and Fees Agency Commissions	Missing some contextual information about the booking history: Markup calculations/percentages Booking Failures
Focus on providing and explaining itinerary data. Read-only without any impact on the itinerary.	Extend usage of the Cockpit Copilot for itinerary actions: add a processing fee or discount, perform a booking change or cancellation.
Basic error handling: When information cannot be found When a prompt is not clear or applicable to the itinerary	Suggestions on prompts that may achieve the intended results when no information is found.
The UI is an individual page that occupies the whole tab when open, making parallel navigation in Customer Care pages impossible.	Present the Cockpit Copilot in a sidebar instead of a full-page view.

 

Usage of the External OpenAI Platform

The Cockpit Copilot leverages OpenAI’s Large Language Model (LLM) through their AI Platform to generate the response to users' questions. TripBuilder sends users' questions along with a structured data object containing booking details to the OpenAI platform.

Personally Identifiable Information (PII) related to travellers is anonymised by default and not sent to OpenAI. This includes:

• First name
• Last name
• Name
• Email
• Gender
• Date of Birth
• Age
• Mobile Phone
• Phone
• Address
• Nationality
• Passport Number
• Passport Expiry Date
• Passport Issuing Country
• External Reference ID
• Travel Agent

 

Anonymisation Process

When a request is made by the Cockpit Copilot, it is anonymised (replaced by random string placeholders) before being sent to OpenAI. The data is then de-anonymised in the response provided to the user.

Example:

1. Real Data used in the itinerary.

Passenger 1:

• First Name: Lina
• Last Name: Nuñez
• Gender: Female
• Birth Date: 1989-06-03
• Age: 34
• Passport Number: 99FBC1
• Passport Expiration Date: 2030-06-03

 

2. Placeholder Data Sent to OpenAI after anonymisation:

Passenger 1:

• First Name: $$$_fb13965f-1d55-4ab3-889e-010ed23c4996
• Last Name: $$$_6e7ccb09-cc17-4058-87cf-da0a1723e8a1
• Gender: $$$_e752fb38-bf6c-4527-a33f-0580866298b5
• Birth Date: $$$_538ac8b5-ca8c-49d2-96b0-0c507ba7048e
• Age: $$$_314fcab0-ab71-4579-9245-a99d4ebfb06a
• Passport Number: $$$_a4268b4e-fe2a-4eee-9996-f91428e5521a
• Passport Expiration Date: $$$_6a59df2b-710a-4c91-b619-579a15019db8

 

3. After de-anonymisation (e.g., if some data needs to be presented to the Cockpit Copilot user as part of a prompt result):
Passenger 1:

• First Name: Lina
• Last Name: Nuñez
• Gender: Female
• Birth Date: 1989-06-03
• Age: 34
• Passport Number: 99FBC1
• Passport Expiration Date: 2030-06-03
•  

Remaining booking data is maintained and used as-is. Details of travel services (title, dates, sales prices, net prices), names, and IDs of agencies and tour operators are exposed to OpenAI.

Example of Data Sent for Itinerary & Travel Services:

• Title
• Travel Dates
• Sales and net prices
• Identifiers (agency, tour operator, itinerary, suppliers)
• Custom remarks or titles.

Please note that if tour operators or travel agents set custom titles and remarks on the booking that include PII (e.g., setting the itinerary title to "US west coast for family Doe"), this data will not be anonymised and will be sent to OpenAI.

Important: Access to the Cockpit Copilot requires acceptance of its underlying terms of service. Specifically: The data is sent to an AI service. Please note that by using our TripBuilder AI Cockpit Copilot, you acknowledge and accept full responsibility for the data you submit, including ensuring its compliance with all applicable laws and regulations. Additionally, be aware that the AI-generated responses may require manual verification and should not be solely relied upon for accuracy.

 

Data Security

Nezasa uses OpenAI's API Platform with their own API keys. The API Platform guarantees the following (quotes from the OpenAI Enterprise Privacy Terms:

• Data Usage and Retention:
  • We do not train on your business data
  • After 30 days, API inputs and outputs are removed from our systems.
  • You own your inputs and outputs.
• Security, Encryption:
  • Enterprise-level authentication through SAML SSO
  • We’ve been audited for SOC 2 compliance
  • Data encryption at rest (AES-256) and in transit (TLS 1.2+)